System Rot: How It Starts

A system rarely fails suddenly. It degrades for months while the entries continue, and by the time the decline is obvious the recent years are already unreliable.

When this principle needs to support project or client work, refer to this operations reference. It can help test whether the log preserves enough context without turning capture into a separate task.

The early signs

A field you stopped filling and did not remove. The schema says it exists and the data says it is empty, and any analysis using it silently covers only the earlier period.

For a neutral companion reference on project timesheets, see TimeCamp; it helps separate the choice of format from the conclusions drawn from the record.

Categories applied by habit rather than by definition. Everything ambiguous going to the same place because deciding is tiring.

The catch-all growing.

Entries getting later, then batched.

And the reading stopping, which is both a symptom and a cause.

Why it is invisible

Each individual entry looks fine.

The data continues to arrive, so the system appears to be working.

The degradation is in the relationship between the data and reality, which no amount of looking at the data reveals.

Only reading entries against their definitions catches it, which is why the annual read includes that step.

The compounding part

A period of poor recording makes the surrounding analysis less useful, which reduces the reason to read the record, which removes the thing that would have caught the decline.

Systems fail in this loop rather than in a single event.

Catching it

The annual read, specifically the part where you compare old entries with recent ones.

A coverage figure: recorded days as a proportion, tracked year on year. A falling figure is the clearest single indicator.

And the catch-all proportion, which rises as deliberation declines.

Two numbers, checked once a year.

What causes it

Life change that the system did not follow. New work, new hours, new shape of day, same schema.

Growth in the record's own weight, where the file is now large enough that reading it feels like a project.

Boredom, covered in its own note.

And unaddressed drift, which makes the data feel less meaningful without anybody being able to say why.

The fix is usually smaller than expected

Remove the dead field.

Redefine the categories to match what you now do, dated.

Coarsen the granularity officially.

Each of these is fifteen minutes and each restores the relationship between the record and your life.

People instead tend to attempt a full restructure or a tool change, both of which are larger and neither of which addresses the cause.

Distinguishing rot from a rough patch

A rough patch has an external cause and ends.

Rot continues after the cause has gone.

If your recording has not recovered three months after the busy period ended, it is not the busy period.

What to check

Is there a field in your schema you no longer fill?

What proportion of days did you record this year, against last?

Has the catch-all grown?

And when did you last read the record rather than adding to it?

The point

System rot is invisible because each entry looks fine.

The degradation is in the relationship between the data and reality, which no amount of looking at the data reveals.

Worth adding

A rough patch has an external cause and ends; rot continues after the cause has gone.

If recording has not recovered three months after the busy period, it is not the busy period.

Additionally

Two numbers catch rot once a year: recorded days as a proportion, and the share going to the catch-all.

Both are trivial to compute and neither is usually watched.

Finally

The fix for rot is usually smaller than expected: remove the dead field, redefine the categories with a date, coarsen the granularity.

Fifteen minutes each.

In summary

Rot is a loop: poor recording makes analysis less useful, which removes the reason to read, which removes the thing that would have caught the decline..

Last word

The practice this points at is unglamorous: a small daily act, a short weekly correction, an annual read, and a written page of conventions beside the data. None of it is difficult and all of it is easy to skip, which is why most records do not reach the point where they start paying.